Seeing the Winds Better: Simulated Videos of Defoliated Tree Motion Capture Extreme Wind Speeds

Extreme windstorms are among the costliest natural disasters in northwest Europe. Traditional wind measurement methods, while reliable, are limited by cost, installation complexity, and sparse spatial coverage, particularly in cluttered urban areas. Higher resolution approaches are therefore needed to monitor near-surface wind dynamics in complex settings.

Motion tracking in videos of foliated trees has reliably captured fine-scale wind variability, showing strong correlations with anemometer measurements, consistent gust estimates across reference objects, and strong temporal coherence; the present work introduces two major advances: (1) the first application of Visual Anemometry (VA) on videos generated from physics-based simulations of trees, and (2) a focus on defoliated trees, enabling mechanistic isolation of branch and trunk responses. Videos are generated from an elastically articulated body model simulating tree responses under mean wind speeds of 7-40 m s⁻¹ (25-144 km h⁻¹).

Results show that input wind speeds are reflected in kinematic tree responses (R² = 0.8), and then that these tree motions are captured in wind speeds estimated from videos by VA (R² ≈ 0.7). Distal and mid-canopy branches dominate the VA response, whereas the stem and inner branches provide only weak contributions, even though informative motion cues may occur anywhere in the canopy. These VA wind estimates were insensitive to camera orientation, confirming that estimation accuracy remains robust across horizontal viewpoints. In this work, we also explore the method's sensitivity to different camera parameters and assess its transferability across different conditions.

Using simulation-based defoliated trees, this work is a step towards a low-cost, scalable alternative to traditional methods, enabling improved detection of extreme gusts, fine-scale hazard mapping, and risk assessment for urban planning and insurance.